Skip to content
Polytope

Writing

How the words on pages built with Polytope should read: plain, specific, and in the reader's order.

The model learns on simplicial complexes, so it can use relations among three or more nodes at once.

Plain over grand. The same claim twice. The second version says what the model does, and every word carries information.

Voice

Write the way you would explain the work to a colleague at the whiteboard: plain words, concrete examples, short sentences mixed with longer ones. Be confident about what the results show and exact about what they do not.

Rules

  • Lead with the point. The first sentence of a page or section says what it is about.
  • State claims directly. Skip the "not just X, but Y" setup and say Y.
  • Use numbers with units and conditions: "0.84 test accuracy on Cora, mean of five seeds", not "strong results".
  • Leave out sales words: groundbreaking, cutting-edge, seamless, robust (outside statistics), novel (let the reader decide).
  • Do not end a section with a one-line summary of what it just said.
  • Use commas, colons, and parentheses. Avoid em and en dashes in prose.
  • Keep bold for words a reader must not miss. A list does not need a bold label on every line.

Interface text

  • Buttons say what happens: "Copy BibTeX", "Run again", "Download CV".
  • Links say where they go: "the paper on arXiv", not "here".
  • Errors name the problem and the fix: "Learning rate must be above 0. Try 0.01."
  • Empty states say what is missing and what to do next.

Research content

  • Paper titles in sentence case on the web, even when the venue uses title case.
  • Author lists in the order of the paper. The site owner's name in bold in publication lists, nowhere else.
  • Dates as "11 October 2026" in prose and "Oct 2026" in lists.
  • Figures get a caption that starts with what the figure shows.
  • Never present example or placeholder content as real. Mark it, or leave the section out until the real content exists.

Spelling and terms

American spelling. Terms as in the field: simplicial complex, cell complex, message passing, Hodge Laplacian, sheaf. Spell out an abbreviation the first time on each page: graph neural network (GNN).